Tehran.
2. Data and Method
We used all SAR data including descending and ascending orbits for the Envisat between 2003 and 2009 and ascending
orbit for ALOS in dual polarization mode between 2007 and 2009 for our study area Fig. 1. We selected this region
because of potential volcanic and landslide hazards around Damavand Volcano. In this paper we focused on 2 landslide
targets located in KAH and EMZ shown in Figure 1 and evaluate the effectiveness of our modified method MSBAS
by comparing the results with those derived from SBAS method implemented in StaMPS Hooper et al., 2004.
Figure 1: Left: Photo of Damavand Volcano, as viewed from east, Right: Deformation field velocity of Envisat satellite as
obtained in descending. The location of the KAHR and EMZ are depicted.
Due the effect of geometric decorrelation on the accuracy of unwrapping process, we used a modified phase filtering with
the aim of smoothing each SBAS pairs before unwrapping. However, despite of using modified filter and because of
strong geometric decorrelation in rough terrain area, some areas still remained noisy. Therefore, we only selected
correlated areas from each SBAS pairs. In this regard, we applied a unique mask for all of the SBAS pairs during 3D
unwrapping and used the designed mask in all unwrapping steps to improve the results. We refined the Delauny
triangulation and removed edges lying beyond masked areas during time unwrapping. We also applied the same mask for
the regularly gridded data that is interpolated from PS pixels during spatial unwrapping.
Due to the sensitivity of interferomatric phase to the digital elevation models DEMs errors, especially for ALOS dataset,
and because the Nyquist condition for time unwrapping will not be met if there are any spatially uncorrelated terms, the
correct detection of the topographic errors in the time-series of interferograms is a critical step. StaMPS software ignores the
topographic correction before unwrapping and only detects the topographic errors after unwrapping process. In StaMPS,
irregularly sampled PS pixels are unwrapped using 3D unwrapping algorithms. The algorithm unwraps time series
pixels in both time and space domain separately. In time domain, the StaMPS algorithm works very well only for well
sampled case in which Nyquist assumption is satisfied, but has problem for rough terrain areas where phase is undersampled
in time due to topographic, atmospheric and noise effects or errors. Therefore, we modified StaMPS 3D unwrapping
algorithm to overcome these problems. Similar to StaMPS, unwrapping in time domain is the first step in our algorithm.
The algorithm defines edges connecting data points using Delaunay triangulation and then calculates edge phase
differences. In this way only spatially correlated effects including atmospheric and orbital errors are removed but there
remain several time undersampled areas due to topographic errors that has to be corrected. StaMPS solution for adjusting
these areas time undersampled areas is to apply the averaging filter in time domain, regardless of the source of
error. In our algorithm, in contrast to StaMPS, topographic correction is performed on the measured edge phase
differences difference map. We reduce the phase component correlated with the perpendicular baselines from each phase
difference map due to the possible topographic noise errors to make the phase difference as small as possible. Our
algorithm also could detect the correlation between topography and phase data due to the tropospheric delay, and
remove it after unwrapping process. In addition, the possible precise orbit errors were removed by fitting a best plane on
four GPS permanent stations available in our study area.
3. Topographic correction method
Let us first consider components constitute interferometric phase. These parameters are the phases introduced by the
topographic effect
topo
, the phase difference due to deformation between the two acquisitions
def
, the atmospheric phase delay
atmo
, orbit errors
orbit
, and noise term
noise
eq. 1.
j
=
j
def
+
j
topo
+
j
orb
+
j
atm
+
j
noise
1 The main objective is to separate phase contributions due to
deformation
j
def
from other terms. Deformation information can be accurately estimated from SAR
observation if a true elevation model DEM of the area is available. Any factor that limits DEM accuracy whether due to
instrument or methods used for topographic measuring and also possible changes in Earths topography during time series
interval could be affect phase unwrapping performance and consequently reduces the derived deformation accuracy.
There are linear relationship between the phase shifts caused by DEM error or topographic changes and perpendicular
baselines Samsonov., 2010 eq. 2.
e e
topo
Z r
B sin
t ,
t 4
= t
, t
j i
j i
2
Where
e topo
is the phase component due to the error in the DEM,
t ,
t
j i
B
is perpendicular baseline between the two time acquisition
i
t
and
j
t
, r is SAR range between SAR sensor and target, θ in the look angle and
e
Z is the DEM
error. We assume DEM error component dominate the measured phase shift, since other terms which uncorrelated in
time including atmospheric, orbital, thermal noise, and so on have already been removed by differencing between nearby
pixels. The pixels with strong correlation between phase and
baselines are
found as pixels which include DEM error. To estimate residual topographic ratio
sin 4
r Z
e
, by taking that in equation 2 we applied linear regression
between the calculated term
e topo
and
B
.
Damavand Volcano
KAH
EMZ
SMPR 2013, 5 – 8 October 2013, Tehran, Iran
This contribution has been peer-reviewed. The peer-review was conducted on the basis of the abstract. 448
4. Results 4.1 Envisat descending results
Kategori